Rarely does a marketing database project fail for purely technical reasons. Most marketing database projects have the hardware and software they need to be successful. So why do so many fail? The three leading reasons are:
- People assume that with better knowledge of their customers, they will do the same things more efficiently. In reality, they end up doing different things. Gaining better knowledge often demands greater changes than the organization can accept and use effectively. As a result, the marketing database becomes a tactical tool of limited value instead of a truly valuable strategic tool.
- The database is built without an understanding of the business model. Simply, a "business model" is a description of how the business behaves to make a profit. Failing to understand the business model dooms many database construction efforts because of choosing the wrong champion, the wrong platform, etc.
- The business sees the marketing database as a thing when in fact is part of a larger process. Too often, a business sets out to build a marketing database, often spending millions of dollars to do so. Nevertheless, the business fails to gain a bottom-line benefit from the investment because having a marketing database is not the goal. Using it is!
THE BUSINESS MODELING PROCESS
The business modeling process is a way to monitor the relationship of the business's strategy and with respect to its environment. A marketing database is at the heart of every successful business modeling process. Why? Because the marketing database is the means by which the top management gains objective knowledge about their changing markets and competitive pressures. It does this primarily by measuring customer and prospect behavior.
A successful marketing databases is a decision support tool that guides both business strategy and day-to-day tactics. It must involve all parts of an organization: top management, marketing, information systems, finance, production, product planning and other departments.
Successful business modeling also requires a desire and willingness to change. Creating a business model or a marketing database does no good unless the business uses it to alter strategies, create tests, track results and react to new or competitive business models. The company must adopt the view that its business strategy and business model is dynamic and responsive to its changing environment instead of being relatively static. Adopting this new perspective can be very challenging!
ESTABLISHING A PROCESS
There is a process for building a useful marketing database. Each step begins and ends with reference to the business model. The process can be broken down into overlapping iterative phases. They include:
- Initial study
- Database marketing
- Data mining
- Customer relationship management (CRM) and
- Corporate integration (see figure below)
The initial study identifies the current business strategy, tactics and resources being applied to the perceived business model. Remember that the perception of the business model is likely to change as better data is acquired. Database marketing organizes the data in a useful format and identifies basic marketing knowledge. Data mining takes a more sophisticated look at the data and describes high-level relationships. Customer Relationship Management begins the regular implementation and use of the knowledge to improve financial performance. Corporate integration is finally achieved when the marketing database is used to support strategic planning.
Building a marketing database cannot be done by the marketing department alone. Expertise and cooperation from many departments is needed. It requires a team of domain experts working together with a clear charter from top management.
Fortunately, there are many early benefits from the process of building a marketing database. These are the result of improvements in corporate knowledge as the project progresses. The following steps outline a process that has proven effective for many industries as they develop their marketing databases.
STEP 1: REVIEW SITUATION ANALYSIS AND BUSINESS MODEL
The goal of this step is to understand the business model so that the marketing database will support the needs of the enterprise. The process begins by developing a basic understanding of the current business model by reviewing the following:
- Current business plan
- Current marketing strategy and tactics (a subset of the business plan)
- Current data structures formats, expected data quality and internal data availability
- The historical period covered by available data
- How revenue is reported, how costs are applied and what reporting periods are used
- Outside marketing threats and opportunities
- Secondary information available outside the company
- Economic environment (both external and internal)
- Comparison of current organization structure to the business plan and business model.
STEP 2: PROPOSE A DATA DICTIONARY AND REQUEST DATA
This step defines the expected data components needed from internal and external data and to begin assembling the marketing database. Specific requests are based upon information uncovered in Step 1 and include:
- Which individuals/departments will provide data
- Which files, fields and records are to be collected
- What documentation is to accompany each file provided
- What are the expected deadlines.
STEP 3: LOAD DATA
As the requested data arrive, they must be assembled into a standard format so they can be manipulated and converted to match the data dictionary. Then a single marketing database can be built. Data seldom arrive in a complete, correct manner on the first attempt so this step may have to be repeated several times. This part of the process includes
- Comparing data arriving to the data requested
- Comparing data documentation against what was received
- Physically loading the data into the system
- Converting data as necessary to a standard, usable database format
- Checking for completeness of data provided by field and record, and
- Re-submitting requests to replace incomplete or unusable data.
STEP 4: CHECK THE DATA FOR SANITY
It is essential that the data used to build the marketing database are accurate and complete. Otherwise, major delays and expense will occur. This requires more than simply checking records and fields. For example, sales amount may be off by a decimal point or sales from a given time period may be omitted. Inspection of the following is appropriate:
- Number of invoices/transactions covering summary and product detail
- Sales dollars in total, by year and by period
- Number of customers in total, by region with purchases and without
- Number of records with blank data in fields expected to be populated
- Number of sales reps, products and offers in the data provided
- Verification that key tables relate as expected by key fields such as customer number, invoice/transaction number, product Number/SKU and Sales Rep.
If inaccuracies are found, request new data and/or revise the data dictionary. Then go back to Step 2 for the affected data.
STEP 5: DATA HYGIENE
Organizing the data into a consistent format and removing "trash" from the data is the next step. Steps needed to achieve thorough data hygiene include:
- Placing each data element in the correct field (often a problem with name/company/address data)
- Parsing names (placing title/first/last names in separate fields)
- Standardizing names, addresses functional titles and other categorical information as required
- Finding, marking and in some cases eliminating duplicate records
- Creating unique ID numbers as required
- Placing data into the data structures in the proposed data dictionary
- Creating exception reports for unusual findings (Go back to Step Two for affected data if required.)
Often, overlay data (data from outside sources) are added to the database by matching contact records in the database. This should be done after cleaning the data and marking duplicates to ensure higher match rates. Doing this will minimize the wasted expense of matching data to duplicated records.
STEP 6: SET UP DATA DICTIONARY
Once the data have been cleaned, it is time to revise the data dictionary to reflect the reality of the database structure. This is always necessary because the delivered data are rarely exactly the same as the data requested or expected. Such revisions include:
- Revising field descriptions and file layouts
- Revising tables and entity relationships as needed
- Setting future update information requests and procedures to match the present needs of the marketing database.
STEP 7: DESCRIBE THE DATA
The first analytical results from a marketing database project are simple descriptions of the data. These help everyone understand what data are stored in the database. They also reveal obvious buying patterns reflected in the database. These descriptive reports include:
- All reports from Step 4 (data sanity check) based upon cleaned, de-duplicated data in the final data structure
- Customer summary reports showing number of buyers by sales dollars, frequency of transaction, dates of transactions and customer type (type/size of business, etc.)
- Definitions of "best" customers (using the 80/20 rule).
- Reports of sales by past marketing efforts based upon data available.
- Calculate and summarize life-to-date sales by customer, customer type and marketing source.
STEP 8: DEFINE RELATIONSHIPS
The next step is to determine how behavior is defined such as revenue levels, loyalty, defection, profit and service category and then to simplify that into variables that can be used in predictive and/or descriptive modeling. This is necessary to determine precisely what is being predicted (the dependent variable) and what is being used to predict (the independent variables). This step includes:
- Expectations for the process of statistical analysis and review
- Description of the relationships to be analyzed
- Definition of what is to be predicted or described
- Definition of dependent and independent variables
STEP 9: PERFORM DESCRIPTIVE ANALYSIS
The second set of analytical results are descriptive statistics that may show patterns within and between each variable. This knowledge is very useful in deciding which modeling technique should be used for subsequent analyses. These reports include:
- Cross-tabulated relationship reports for all major variables;
- Confirmation of relationships, or a lack of relationships between major variables
- Demonstration of which relationships are strongest and most likely to be predictive and /or descriptive.
STEP 10: PERFORM PREDICTIVE MODELING
After all of the preparatory work described above, it is finally time to create revenue and loyalty models for customers and potential customers. The models need to provide sufficient descriptive modeling to provide plain-English descriptions of how customers are different. The results from these models typically are extremely helpful in guiding strategic decisions. These results include:
- Predictive modeling for selection/de-selection of customers and potential customers in future communication efforts
- Descriptive model reports showing differences and similarities among customers and potential customers
- Recommendation on how to best use the knowledge discovered in the modeling process.
STEP 11: DEFINE FINAL REPORTS & UPDATES
After models have been built and analyzed, it is time to create the final report set for the initial build and define the report set for future updates. Once the models reveal what variables are most useful for guiding key decisions, this step defines future outputs to provide updates for the most useful variables. These reports include:
- Reports created and revised through Step 10
- Procedures for future updates, including timing and data requests, and
- Develop an ongoing procedure to update and revise the database
STEP 12: DEVELOP & IMPLEMENT A STRATEGY
Next it is appropriate to review findings, discuss the implications and set up a strategy to implement CRM. This step communicates the information available and gives decision makers a chance to revise strategy based upon new knowledge. Normally, major changes are implemented in communication efforts at this point. Results from this step include:
- Presentation and explanation of the final report.
- Review of implications for marketing.
- Discussions of how to best use the data, changes to make immediately and long-term changes to make.
STEP 13: ADJUST & UPDATE THE DATABASE
The adjustment period provides a way to monitor how the marketing database is being used, receive feedback based upon the experiences of users and make changes as required. A marketing database is a complex and powerful tool and only with experience can a business become aware of its strong and weak points. It may seem that not much happens during this part of the project but it is still important. It is a phase-in period, prior to creating an automated, preset data action and reporting mechanism that will be difficult and expensive to change.
The adjustment period typically spans several update cycles so that users have time to see the results, react to them, monitor outcomes and request needed changes. During the adjustment period, the marketing database structure (as defined in the data dictionary) is kept as simple and flexible as possible to allow for changes. Results from this step include:
- Completion of several update and reporting cycles
- Phasing-in greater use of the knowledge derived from the database
- Review of database usage and results
- Review of tracking capabilities of the marketing database to be sure results are being measure and
- Changes to the marketing database and reports where practical and feasible.
STEP 14: DEVELOP APPLICATIONS
The final part of a database marketing development project is to develop data access tools that marketers can easily use to automate as many routine updates, reports and queries as possible. This step places the knowledge and power of the marketing database in the hands of the end-user.
It is important that this be the last step of a project. Defining and designing interfaces and reports before being certain of what data are needed will result in disappointment with the final product, expensive changes or both. Only after the data structure and outputs have proven their value should software application developers create a final product for the client. (The data-access tool may reside anywhere -- on the web, on a service provider's system or on the client's network.)
The results from this final step 14 include:
- A proven set of requirements for final implementation
- A defined data set and update procedure
- A useful set of reports and established data query needs and
- A data-access tool available for direct client use.
APPLYING THE BUSINESS MODELING PROCESS
Support from top-level management is essential to achieving a successful marketing database project. Nevertheless, the effort often begins within a single department. The department's challenge is often gaining the acceptance and cooperation of the rest of the organization.
One way to achieve corporate buy-in is through a "fast start" approach. Instead of undertaking a major development effort, it is usually better to start out with a pilot program. The pilot program can accomplish just a portion of the first few steps, including the initial study and "sanity checks" on samples of real data. It will still be useful and provide real benefit to the business.
There are at least two advantages of the "fast start" approach. The first is that it shows other corporate managers what can be achieved. Second, it establishes how large and how much of a challenge the overall project will likely be. This information is especially useful in setting goals, budgets, timelines and overall expectations.
Unlike the traditional approach to business modeling in which someone (usually an outside consultant) conducts an infrequent study and then moves on to other projects, the transaction-driven business modeling process described in this article requires ongoing assistance. Remember, it is a process, not an event. So, much of the effort to implement and use a marketing database must come from within the enterprise. Still, it is still wise to include the viewpoint of a specialist throughout the process.
Questions? Just ask! [Back to How-to Resources] [Top] |
